Atlas

Benchmarks

← All benchmarks

Finance Agent v2 - Precedents

Professional Work · 2026-05-19

The Precedents split of Finance Agent v2. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.

Top models (higher is better)

ModelScore
Gemini 4 Argon49.8
Muse Spark Open40.8
Gemini 3.5 Flash36.4
Gemini 3.8 Flash36.1
Muse Spark 1.235.4
Muse Spark 1.335.2
Muse Wonder Full35.0
Gemini 3.7 Flash33.8
Muse Spark 1.133.8
Claude Fable 5.132.2
Muse Wonder32.1
MiMo-V2.6-Pro30.9
Gemini 3.6 Flash30.8
Hy4 Preview30.6
GPT-6 Astra30.3
MiMo-V2.6-Flash30.3
Qwen3.8 Max (0902)29.5
Ember-129.2
GPT-5.6 Luna29.1
Sonnet 528.8
GLM-5.3 Flash28.6
GPT-6 Sol28.5
DeepSeek V4.1 Flash28.2
GPT-6 Luna27.4
Claude Haiku 5.527.0
DeepSeek V4.1 Flash 101726.5
GPT-5.6 Sol26.2
Mistral Large 4 Preview26.2
Gemini 3.5 Flash-Lite26.0
Grok 4.725.9
GPT-6.1 Sol25.8
Opus 4.825.6
GLM-5.325.1
GPT-5.6 Terra24.8
Inkling24.8
DeepSeek-V4-Flash-073124.1
Kimi K323.6
MiniMax M323.2
GPT-5.4 Mini22.1
Grok 4.621.9
Loading Atlas data…